# Easiest way to count occurrences of DMatch in a vector<DMatch>

Hi,

I have a vector of matches (`vector<DMatch>`

) in which are multiple same matches. Now I want to get the best (so the match with the most occurences) out of this vector.

For understanding:
I calculate a Fundamental matrix a lot of times for the same image pair but the calculation is everytime slightly different. I always save the best matches (most of the times: 7 per iteration) in my `vector<DMatch>`

. Here is the code snippet:

```
vector<DMatch> all_inlier_matches;
// This part is in a for loop with several iterations
vector<DMatch> good_matches, good_matches_2, sym_matches;
BFMatcher matcher;
vector<DMatch> matches, matches_2;
float ratio_Lowe, ratio_Lowe_2;
BruteForceMatching(descriptors_1, descriptors_2, good_matches, matches, ratio_Lowe);
BruteForceMatching(descriptors_2, descriptors_1, good_matches_2, matches_2, ratio_Lowe_2);
symmetryTest(matches, matches_2, sym_matches); // Calculation of the best matches
for (size_t i = 0; i < sym_matches.size(); i++) {
all_inlier_matches.push_back(sym_matches[i]); // Push_back ~seven matches per iteration;
}
```

Now I want the best of my best matches, so the matches which occur for example more than 5 times. So what is the structure of a DMatch, how can I compare them to each other and how can I count same occurences? I read something about counting in a vector with unordered map but I can hardly understand it.

please have a look at the DMatch class

what would you consider "same" here ? like same indices, but distance differs ? are matches with swapped indices "same" ?

You are right. I saw now that you can't really compare the matches itself so I extracted the coordinates of the matched points in two

`vector<Point2f>`

. Now I can just compare my left image point coordinates with my right image point coordinates. When I have equal coordinates on different positions in the vectorfor both imagesI have the same match.